nnsvs.util

General utility

Example files

example_xml_file

Get the path to an included xml file.

Initialization

init_weights	Initialize network weights.
init_seed	Initialize random seed.

Multi-stream helper

get_world_stream_info

Get stream sizes for WORLD-based acoustic features

Mask

make_pad_mask	Make mask tensor containing indices of padded part.
make_non_pad_mask	Make mask tensor containing indices of non-padded part.

Padding

pad_2d

Pad a 2d-tensor.

Path

load_utt_list

Load a list of utterances.

Scalers

class nnsvs.util.PyTorchStandardScaler(mean, scale)

PyTorch module for standardization.

Parameters:

• mean (torch.Tensor) – mean

• scale (torch.Tensor) – scale

class nnsvs.util.StandardScaler(mean, var, scale)

sklearn.preprocess.StandardScaler like class with only transform functionality

Parameters:

• mean (np.ndarray) – mean

• var (np.ndarray) – variance

• scale (np.ndarray) – scale

class nnsvs.util.MinMaxScaler(min, scale, data_min=None, data_max=None, feature_range=(0, 1))

sklearn.preprocess.MinMaxScaler like class with only transform functionality

Parameters:

• min (np.ndarray) – minimum

• scale (np.ndarray) – scale

• data_min (np.ndarray) – minimum of input data

• data_max (np.ndarray) – maximum of input data

• feature_range (tuple) – (min, max)

Misc

dynamic_import

Dynamic import